What the four words mean
Start here if the vocabulary is the problem, because most of the disagreement in this subject is about labels rather than about work. GEO is the one with an origin: it was named in a 2023 arXiv paper, later published at KDD 2024. AEO and LLM SEO arrived through practitioner writing with no defining document behind either. Wikipedia's entry records that no consensus definition separating the terms had been established in the academic literature as of early 2026.
So anyone drawing a firm boundary between them is describing a service rather than a distinction. The four pieces below define each term and then test it against something checkable, which is more useful than arguing about which one is correct.
- the definition of GEO, where the term came from and what Google publishes about it. Covers the two retrieval mechanisms Google names, the four eligibility requirements a page has to meet, plus the two things Google states you do not need.
- whether AEO is a different discipline from GEO, tested on the UK search results. The two queries return eight organic results each and share exactly one domain between them, which settles the question by counting rather than by argument.
- what the third label adds, plus whether the pages teaching it get cited themselves. All eight UK pages ranking for the term held zero AI citations at page level on 10 August 2026, while the domains beneath them held 8,753.
- what AI visibility means, the five ways it gets measured and what each one misses. Includes a worked example of two honest measurements of the same brand disagreeing completely, because they have no common unit.
How much of this is not SEO
About half, on the only measurement we have run. Across twelve UK search results on 10 August 2026, 46 of the 85 sources Google's AI Overview cited did not appear in the organic top ten below it. That is 54%. Counted the other way, 58 of 99 organic results had no citation in the AI Overview sitting above them.
That single figure is the reason the vocabulary exists. Ranking on page one does not get you into the answer above it. Missing page one does not keep you out of it either.The full count, the twelve queries it ran on and the three limits that bound itis the piece to read before anyone quotes an overlap percentage at you, including this one.
What decides whether a model quotes you
Not crawler access, which is where most of the advice in this category stops. Ten UK agencies hold the first page of Google for the money term in this market. On 10 August 2026 every one of them allowed every OpenAI crawler. Their ChatGPT citation counts ran from 0 to 92.
Access is the entry ticket rather than the race.The four OpenAI crawlers and what blocking each one actually costscovers the part that does move it, along with the Cloudflare setting that changed in July 2026 and takes sites out of ChatGPT search without anyone noticing.
How to check where you stand
Before buying anyone's dashboard, run the measurement by hand. Fifteen buying questions, three runs each, across the surfaces your buyers use, recorded in four columns. Seventy-five rows is enough to see the pattern. The step that matters is reading the citations rather than the answers, because the same small set of pages will be doing most of the work and that set is what you have to change.
The method, the prompt shapes and the three reasons you are missingtakes an afternoon and costs nothing. It is also the only way to know whether a paid tool would tell you something you did not already know.
Our own number, published from zero
Zero. On 10 August 2026 findable.digital recorded no citations across ChatGPT, Perplexity, Google AI Overviews, Gemini, Copilot or Grok. The site had no articles on it until that month. We publish that table again every month from the same baseline, whichever way it moves, because an agency selling AI visibility that will not show its own is asking you to take the one thing it can prove on trust.
What this pillar does not cover yet
7 pieces is not the subject. The gaps are named here rather than left for you to find. Each one is planned rather than dropped.
- How Google assembles an AI Overview. The retrieval and fan-out mechanisms are described in outline inside the GEO definition. There is no page here taking the mechanism apart properly.
- The ranking factors that appear to matter in ChatGPT. What exists today is the crawler and citation-concentration evidence, not a factor list.
- Structured data and page structure for citation. Nothing published. This is the most requested and the least evidenced part of the subject, which is why it is not written yet.
- Tracking citations over time, plus reading AI traffic in GA4. The five measurement methods are covered. The mechanics of running one monthly, or of separating AI Overview traffic inside analytics, are not.
- What a citation is worth against a click. The comparison is asserted across several of these pieces. Nobody here has costed it, including us.
One limit applies to everything above. Every figure on this page and in the pieces beneath it comes from a single pull on 10 August 2026. AI citation sets have been observed to change within a single day. Treat all of it as the reading on that date rather than as a constant.